How to Use Pickle for Model Saving in Python with sklearn

Use pickle.dump() to save a trained sklearn model to a file and pickle.load() to load it back. This lets you store your model on disk and reuse it later without retraining.

Syntax

To save a model, use pickle.dump(model, file). To load it back, use model = pickle.load(file). The file should be opened in binary mode.

• model: your trained sklearn model object
• file: a file object opened with open(filename, 'wb') for saving or open(filename, 'rb') for loading

python

import pickle

# Save model
with open('model.pkl', 'wb') as f:
    pickle.dump(model, f)

# Load model
with open('model.pkl', 'rb') as f:
    model = pickle.load(f)

Example

This example trains a simple sklearn logistic regression model on sample data, saves it using pickle, then loads it back and makes a prediction.

python

from sklearn.datasets import load_iris
from sklearn.linear_model import LogisticRegression
import pickle

# Load data
iris = load_iris()
X, y = iris.data, iris.target

# Train model
model = LogisticRegression(max_iter=200)
model.fit(X, y)

# Save model
with open('iris_model.pkl', 'wb') as f:
    pickle.dump(model, f)

# Load model
with open('iris_model.pkl', 'rb') as f:
    loaded_model = pickle.load(f)

# Predict with loaded model
sample = X[0].reshape(1, -1)
prediction = loaded_model.predict(sample)
print(f"Predicted class: {prediction[0]}")

Output

Predicted class: 0

Common Pitfalls

• Forgetting to open the file in binary mode ('wb' for writing, 'rb' for reading) causes errors.
• Pickle files are Python-specific and may not be secure to load from untrusted sources.
• Changing the model code or sklearn version may cause incompatibility when loading saved models.
• Always test loading and prediction after saving to ensure the model works as expected.

python

import pickle

# Wrong: opening file in text mode causes error
# with open('model.pkl', 'w') as f:
#     pickle.dump(model, f)  # This will raise an error

# Right way:
with open('model.pkl', 'wb') as f:
    pickle.dump(model, f)

Quick Reference

Remember these key points when using pickle for model saving:

• Use open(filename, 'wb') to save and open(filename, 'rb') to load.
• Use pickle.dump(model, file) to save the model.
• Use model = pickle.load(file) to load the model.
• Test loaded models before using in production.

Key Takeaways

Always open files in binary mode when saving or loading with pickle.

Use pickle.dump() to save and pickle.load() to load sklearn models.

Test your saved model by loading and predicting to ensure it works.

Avoid loading pickle files from untrusted sources for security reasons.

Model compatibility may break if sklearn or model code changes after saving.